Neural network based nonlinear MU-MIMO precoding

Abstract

A base station may apply a nonlinear precoding to data for MU-MIMO transmission to a set of paired UEs to generate a first set of precoder symbols, and apply a linear precoding to the first set of precoder symbols to generate a second set of precoder symbols using a linear precoding matrix. The base station may normalize the second set of precoder symbols, and scale the second set of precoder symbols, before transmission of the data, using a scaling factor based on one or more of modulation symbols or a channel matrix. The base station may apply the linear precoding to DMRS associated with the data. The base station may transmit the second set of precoder symbols based on the second set of precoder symbols and the DMRS to the set of paired UEs.

What is claimed is: 
     
       1. A method of wireless communication at a network entity, comprising:
 applying a nonlinear precoding to data for multi-user multiple input multiple output (MU-MIMO) transmission to a set of paired user equipment (UEs) to generate a first set of precoder symbols; 
 applying a linear precoding to the first set of precoder symbols to generate a second set of precoder symbols using a linear precoding matrix; and 
 transmitting the second set of precoder symbols to the set of paired UEs based on the second set of precoder symbols, 
 wherein the nonlinear precoding is based on one or more of a plurality of modulation and coding schemes (MCSs) for the set of paired UEs, a channel matrix for the set of paired UEs representing channel propagation information between the network entity and the set of paired UEs, or a channel correlation matrix for the set of paired UEs. 
 
     
     
       2. The method of  claim 1 , wherein the nonlinear precoding is performed by a neural network or a machine learning model. 
     
     
       3. The method of  claim 2 , wherein the nonlinear precoding is performed by a hypernetwork type neural network having a last layer that outputs a precoding matrix. 
     
     
       4. The method of  claim 2 , wherein the nonlinear precoding is performed by the neural network that outputs the first set of precoder symbols provided as an input to the linear precoding. 
     
     
       5. The method of  claim 1 , wherein the nonlinear precoding is further based on a plurality of modulation symbols to be transmitted to the set of paired UEs. 
     
     
       6. The method of  claim 1 , further comprising:
 scaling the second set of precoder symbols, before transmission of the data, using a scaling factor based on the plurality of modulation symbols or the channel matrix. 
 
     
     
       7. The method of  claim 6 , further comprising:
 generating the scaling factor based on a neural network. 
 
     
     
       8. The method of  claim 6 , further comprising:
 generating the scaling factor based on one or more of the channel matrix for the set of paired UEs or the plurality of modulation symbols for the set of paired UEs. 
 
     
     
       9. The method of  claim 6 , further comprising:
 normalizing an output of the linear precoding prior to the scaling. 
 
     
     
       10. The method of  claim 1 , further comprising:
 scaling the linear precoding matrix such that a long-term average based on an expectation operation is within a threshold average transmit power. 
 
     
     
       11. The method of  claim 1 , wherein the first set of precoder symbols comprises N vectors, N being a number of streams for the set of paired UEs. 
     
     
       12. The method of  claim 1 , further comprising:
 applying the linear precoding to a demodulation reference signal (DMRS) associated with the data based on a channel between the network entity and the set of paired UEs; and 
 transmitting the DMRS to the set of paired UEs. 
 
     
     
       13. The method of  claim 12 , wherein the DMRS is transmitted without normalization or scaling that is applied to the data. 
     
     
       14. The method of  claim 1 , wherein the second set of precoder symbols is comprised in a physical downlink shared channel (PDSCH) transmission. 
     
     
       15. An apparatus for wireless communication at network entity, comprising:
 a memory; and 
 at least one processor coupled to the memory and configured to:
 apply a nonlinear precoding to data for multi-user multiple input multiple output (MU-MIMO) transmission to a set of paired user equipment (UEs) to generate a first set of precoder symbols; 
 apply a linear precoding to the first set of precoder symbols to generate a second set of precoder symbols using a linear precoding matrix; and 
 transmit the second set of precoder symbols to the set of paired UEs based on the second set of precoder symbols, 
 
 wherein the nonlinear precoding is based on one or more of a plurality of modulation and coding schemes (MCSs) for the set of paired UEs, a channel matrix for the set of paired UEs representing channel propagation information between the network entity and the set of paired UEs, or a channel correlation matrix for the set of paired UEs. 
 
     
     
       16. A method of wireless communication at a user equipment (UE), comprising:
 receiving a multi-user multiple input multiple output (MU-MIMO) data transmission from a network entity; 
 decoding the MU-MIMO data transmission in part based on a linear precoding; and 
 decoding the MU-MIMO data transmission in part based on a nonlinear precoding, 
 wherein the nonlinear precoding is based on one or more of a plurality of modulation and coding schemes (MCSs) for a set of paired UEs including the UE, a channel matrix for the set of paired UEs representing channel propagation information between the network entity and the set of paired UEs, a channel correlation matrix for the set of paired UEs, or modulation symbols to be transmitted to the set of paired UEs. 
 
     
     
       17. The method of  claim 16 , wherein the decoding based on the nonlinear precoding is performed by a neural network or a machine learning model. 
     
     
       18. The method of  claim 16 , wherein the UE decodes the MU-MIMO data transmission using a channel estimation of a precoded channel based on a linear precoding matrix between the network entity and the UE. 
     
     
       19. The method of  claim 16 , further comprising:
 receiving a demodulation reference signal (DMRS) associated with the MU-MIMO data transmission and precoded based on the linear precoding; and 
 performing a channel estimation based on the DMRS to determine a precoded channel. 
 
     
     
       20. The method of  claim 19 , wherein the DMRS is received without normalization or scaling that is applied to the data. 
     
     
       21. The method of  claim 16 , wherein the MU-MIMO data transmission comprises a physical downlink shared channel (PDSCH) transmission. 
     
     
       22. An apparatus for wireless communication at user equipment (UE), comprising:
 a memory; and 
 at least one processor coupled to the memory and configured to:
 receive a multi-user multiple input multiple output (MU-MIMO) data transmission from a network entity; 
 decode the MU-MIMO data transmission in part based on a linear precoding; and 
 decode the MU-MIMO data transmission in part based on a nonlinear precoding, 
 
 wherein the nonlinear precoding is based on one or more of a plurality of modulation and coding schemes (MCSs) for a set of paired UEs including the UE, a channel matrix for the set of paired UEs representing channel propagation information between the network entity and the set of paired UEs, a channel correlation matrix for the set of paired UEs, or modulation symbols to be transmitted to the set of paired UEs. 
 
     
     
       23. The apparatus of  claim 22 , wherein the decoding based on the nonlinear precoding is performed by a neural network or a machine learning model. 
     
     
       24. The apparatus of  claim 22 , wherein the UE decodes the MU-MIMO data transmission using a channel estimation of a precoded channel based on a linear precoding matrix between the network entity and the UE. 
     
     
       25. The apparatus of  claim 22 , wherein the at least one processor is further configured to:
 receive a demodulation reference signal (DMRS) associated with the MU-MIMO data transmission and precoded based on the linear precoding; and 
 perform a channel estimation based on the DMRS to determine a precoded channel. 
 
     
     
       26. The apparatus of  claim 22 , wherein the MU-MIMO data transmission comprises a physical downlink shared channel (PDSCH) transmission.